Agentic Coding Design Patterns

Agentic Coding Design Patterns

Skill by ara.so — Design Skills collection.

A comprehensive catalog of patterns and anti-patterns for working effectively with AI coding agents, inspired by the "Gang of Four" design patterns. This project documents proven approaches for task setting, context management, verification, and project organization when collaborating with AI.

What This Project Does

The Agentic Coding Design Patterns catalog provides:

• Structured patterns for AI-assisted development workflows
• Anti-patterns to avoid common pitfalls
• Multi-language support (English, Russian, Spanish)
• PDF/EPUB export for offline reading
• Pattern categories: task setting, context management, verification, project organization

Built with HonKit as a multi-language book with print capabilities.

Installation

# Clone the repository
git clone https://github.com/mokevnin/agentic-coding-design-patterns.git
cd agentic-coding-design-patterns

# Install dependencies
npm install

Key Commands

Development & Preview

# Start local development server with live reload
npm start

# Build static site to ./dist
npm run build

# Generate PDF for a specific locale (requires Calibre/ebook-convert)
npm run pdf:ru
npm run pdf:en
npm run pdf:es

Project Structure

book/                 # Book root (HonKit)
  book.json           # HonKit configuration
  LANGS.md            # Language list for multilingual support
  <locale>/           # Language-specific content (ru, en, es)
    SUMMARY.md        # Table of contents (groups = headings, patterns = links)
    README.md         # Locale landing page
    <slug>.md         # Individual pattern chapters (flat structure)
  assets/<slug>/      # Shared diagrams and code samples
templates/<locale>/   # Localized pattern and anti-pattern templates

Configuration

HonKit Configuration (book/book.json)

{
  "title": "Agentic Coding Design Patterns",
  "language": "en",
  "plugins": [
    "theme-default",
    "-sharing",
    "sitemap",
    "multilingual"
  ],
  "pluginsConfig": {
    "sitemap": {
      "hostname": "https://mokevnin.github.io/agentic-coding-design-patterns/"
    }
  }
}

Language Configuration (book/LANGS.md)

* [Русский](ru/)
* [English](en/)
* [Español](es/)

Pattern Structure

Creating a New Pattern

Each pattern follows a consistent markdown structure:

# Pattern Name

## Context
Description of when this pattern applies

## Problem
The challenge this pattern addresses

## Solution
How to implement the pattern

## Example
Concrete code or workflow examples

## Consequences
Benefits and trade-offs

## Related Patterns
Links to similar or complementary patterns

Pattern Template Usage

# Use locale-specific templates
templates/en/pattern.md        # English pattern template
templates/en/anti-pattern.md   # English anti-pattern template
templates/ru/pattern.md        # Russian pattern template

Adding a Pattern to Table of Contents

Edit book/<locale>/SUMMARY.md:

# Summary

* [Introduction](README.md)

## Task Setting Patterns
* [Clear Scope Definition](clear-scope-definition.md)
* [Incremental Task Breakdown](incremental-task-breakdown.md)

## Context Management Patterns
* [Context Window Optimization](context-window-optimization.md)

## Anti-Patterns
* [Over-Prompting](over-prompting.md)

Working with Translations

The canonical locale is Russian (ru) — patterns are first written in Russian, then translated to English and Spanish.

Translation Workflow

1. Create pattern in book/ru/<slug>.md
2. Add entry to book/ru/SUMMARY.md
3. Translate to book/en/<slug>.md (same slug)
4. Translate to book/es/<slug>.md (same slug)
5. Update SUMMARY.md for each locale
6. Share assets in book/assets/<slug>/

Referencing Shared Assets

<!-- In any locale, reference shared assets -->
![Diagram](../assets/pattern-name/diagram.png)

```python
# Reference shared code samples
# See: ../assets/pattern-name/example.py

Common Patterns for AI Agent Collaboration

1. Task Setting Patterns

Clear Scope Definition: Define explicit boundaries, success criteria, and constraints before starting work.

Example task:
- Objective: Implement user authentication
- Scope: Email/password only, no OAuth
- Constraints: Use existing DB schema, <100 LOC
- Success: Tests pass, PR approved

Incremental Task Breakdown: Split large tasks into verifiable increments.

Phase 1: Database migration
Phase 2: Backend API endpoints
Phase 3: Frontend integration
Phase 4: Tests and documentation

2. Context Management Patterns

Context Window Optimization: Provide relevant context without overwhelming the agent.

# Good: Targeted context
git diff main..feature -- src/auth/

# Avoid: Dumping entire codebase
cat **/*.py | pbcopy

Selective File Inclusion: Include only files relevant to the current task.

Focus files:
- src/auth/login.py (modify)
- tests/test_auth.py (add tests)
- docs/api.md (update)

3. Verification Patterns

Incremental Validation: Verify each step before proceeding.

# After each change
npm test
npm run lint
git diff --check

Explicit Acceptance Criteria: Define testable success conditions.

Acceptance criteria:
- [ ] All tests pass
- [ ] No linting errors
- [ ] API response time < 200ms
- [ ] Documentation updated

4. Project Organization Patterns

Flat Pattern Structure: Keep patterns as flat files for easy GitHub browsing.

book/en/
  pattern-one.md
  pattern-two.md
  anti-pattern-one.md

Localized Templates: Provide consistent structure across languages.

cp templates/en/pattern.md book/en/new-pattern.md
# Edit book/en/new-pattern.md
# Add to book/en/SUMMARY.md

Anti-Patterns to Avoid

Over-Prompting

Providing excessive detail that confuses rather than clarifies.

Context Dumping

Sharing entire codebases without filtering for relevance.

Vague Requirements

Asking for features without clear success criteria or constraints.

Skipping Verification

Moving to the next task without validating the current output.

Troubleshooting

HonKit Build Fails

# Clear cache and rebuild
rm -rf _book dist
npm run build

PDF Generation Issues

# Ensure Calibre is installed
# macOS
brew install --cask calibre

# Ubuntu/Debian
sudo apt-get install calibre

# Then generate PDF
npm run pdf:en

Missing Translations

If a pattern exists in ru/ but not en/:

# Check SUMMARY.md for missing entries
diff book/ru/SUMMARY.md book/en/SUMMARY.md

# Copy and translate
cp book/ru/pattern.md book/en/pattern.md
# Edit book/en/pattern.md for translation

Live Reload Not Working

# Kill existing HonKit process
pkill -f honkit

# Restart
npm start

Contributing Patterns

See CONTRIBUTING.md for detailed guidelines on:

• Pattern naming conventions (slug format)
• Required sections for patterns
• Translation workflow
• Asset organization
• Pull request process

Resources

• Online Documentation: https://mokevnin.github.io/agentic-coding-design-patterns/
• Repository: https://github.com/mokevnin/agentic-coding-design-patterns
• License: Text/diagrams under CC BY-SA 4.0, code samples under CC0

Example: Adding a New Pattern

# 1. Create pattern in canonical locale (Russian)
cp templates/ru/pattern.md book/ru/efficient-prompting.md

# 2. Edit the pattern
vim book/ru/efficient-prompting.md

# 3. Add to Russian table of contents
vim book/ru/SUMMARY.md
# Add: * [Efficient Prompting](efficient-prompting.md)

# 4. Translate to English
cp templates/en/pattern.md book/en/efficient-prompting.md
vim book/en/efficient-prompting.md

# 5. Update English TOC
vim book/en/SUMMARY.md

# 6. Preview changes
npm start

# 7. Build and verify
npm run build

This catalog helps AI coding agents understand and apply proven patterns for effective human-AI collaboration in software development.